The most practical method of assessing patient compliance is to ask the patient for a self-report; but, how valid is such a report? This study compared one particular interview question to a pill-count measure by studying 98 patients who visited their family physician, received medication instructions, and were interviewed in their homes ten days later. The interview question identified 80 percent of the true noncompliers, as assessed by the pill count (sensitivity). However, the interview question was not equally sensitive for all subgroups of patients. Results for younger patients, patients receiving renewals rather than new prescriptions, and patients taking more than one drug were not as accurate, showing sensitivities of 75 percent, 40 percent, and 60 percent, respectively.